Toward assimilation of observation-derived mixing heights to improve atmospheric tracer transport models

Common transport models use the mixing height (MH) to determine turbulent coefficients and to obtain tracer concentrations in the planetary boundary layer (PBL). We conducted a pseudo data experiment to elucidate the impact of assimilating MHs to improve CO₂ transport within the Stochastic Time-Inverted Lagrangian Transport model (STILT). Transport of CO₂ was simulated for August 2006 with a receptor located at Bialystok, Poland. STILT was driven by meteorology obtained from the Weather Research and Forecasting (WRF) model, using the Yonsei University (YSU) and Mellor-Yamada-Janjic (MYJ) PBL parameterizations, which differ substantially in the produced MHs. To quantify model-data mismatch in CO₂ to errors in vertical mixing, we defined the WRF-YSU simulation as known truth. Pseudo MH observations were sampled from WRF-YSU at locations of real radiosonde stations. These point observations were interpolated in space-time to the entire WRF domain using kriging with an external drift, which combines observed and modeled MHs to create a "best guess" MH field. We prescribed MHs in STILT driven by WRF-MYJ winds with the best guess to study the impact on CO ₂ concentrations. Differences in CO₂ between the STILT simulations were on the order of ̃0-1 and ̃1-10 ppm on average (i.e., bias), with standard deviations of ̃1-3 and ̃4-14 ppm (random error) during day (12 UTC) and nighttime (0 UTC), respectively. These were reduced when using STILT with the best guess (̃50%-80% of the bias, ̃10%-20% of the random error). Simulated CO₂ concentrations and MHs were also compared to measurements made at the Bialystok tall tower.